Kohler maintains that actual scientific practice in fact supports the Kohler strategy. Exhibit A in his argument is a recent article (Kim et al. 2014) detailing part of the mechanism for visual motion detection. Kohler reproduces a figure depicting their model, and argues that the inclusion of the distributed network in the model suggests that the authors are strictly speaking attributing function to the whole system as depicted:
Although it is true that investigators sometimes refer loosely to local elements as displaying a certain characteristic, the corresponding detailed and extended accounts of direction selectivity give credit to the distributed nature of the relevant systems that figure in explanations. (Kohler this collection, p. 6)
I agree that this is one possible interpretation of the practice. But here is another: these scientists are distinguishing between the system that exhibits the phenomenon and the mechanism that produces it, and are open to different sorts of relationships between them. Consider the following from the paper Kohler discusses:
Research on [the visual detection of motion] has converged upon the SAC. An SAC dendrite is more strongly activated by motion outward from the cell body to the tip of the dendrite, than by motion in the opposite direction. Therefore an SAC dendrite exhibits DS, and outward motion is said to be its ‘preferred direction’. Note that it is incorrect to assign a single such direction to a SAC, because each of the cell’s dendrites has its own preferred direction. DS persists after blocking inhibitory synaptic transmission, when the only remaining inputs to SACs are BCs, which are excitatory. As the SAC exhibits DS but its BC inputs exhibit little or none, DS appears to emerge from the BC–SAC circuit. (Kim et al. 2014, p. 331; emphases added)
Far from seeming loose, the attribution of direction-selectivity to the dendrite appears to me clear and precise. Moreover, note that in the final sentence quoted above, the attribution of direction-selectivity to the cell is reinforced, even in the context of a reference to the mechanism as the “BC-SAC circuit”. Indeed, I would argue it is natural and permissible to gloss the last clause in the following way: “DS in the dendrite appears to emerge from the BC-SAC circuit.” On this reading, of course, the authors of this article would be proposing an R3 functional relationship such that parts of the mechanism are on a higher level of organization than the system exhibiting the phenomenon.
That these authors are open to R3 relationships of various sorts appears to be reinforced by a line later in the paper:
Previous research suggests that On–Off direction-selective ganglion cells inherit their DS from SAC inputs owing to a strong violation of Peters’ rule. (Kim et al. 2014, p. 335; emphasis added)
Here again we see the same pattern: a clear attribution of direction-selectivity to the DSGC in the same sentence as a reference to the distal mechanism (the SACs), in the context of what is obviously an R3 relationship between system and mechanism. Thus, while I agree that the Kohler strategy is viable, I don’t see that consideration of scientific practice forces us to adopt it, or even necessarily favors it.
So what might be other reasons for adopting the Kohler strategy over extending mechanism to include enabling constraints? As I mentioned at the end of the previous section, the matter might come down to how closely one thinks the architectural facts about the brain match the guiding assumptions of the componential mechanist framework. If one expects that the brain is at root a decomposable or nearly-decomposable system of well-defined interacting components, then componential mechanism does indeed seem like a very appropriate framework for capturing at least the majority of its functional relationships (with the few exceptions to be dealt with perhaps as secondary elaborations or special cases). If, however, one takes seriously the notion that the brain is a massive network marked by multiple, nested, cross-cutting, dynamic hierarchies interacting in bottom-up, top-down, feed-forward and feedback fashions (Pessoa 2014), then one might wish for some of the explanatory flexibility that the notion of enabling constraints appears to offer. I, of course, am in this latter camp (Anderson 2015).